The machine coding round decides most Flipkart loops. Practise building working, extensible code in 90 minutes.
Flipkart popularised the machine coding round in India and still weights it heavily. You get 90 minutes and a problem statement - a parking lot, a splitwise clone, an inventory service - and must produce working, runnable code with sensible class boundaries, then defend it in a follow-up design discussion.
The trap is optimising for output volume. Interviewers do not reward the most features; they reward code where adding the next requirement is obviously easy. Candidates who spend the first ten minutes on entity boundaries and interfaces routinely beat candidates who start typing immediately and finish more.
The design rounds are e-commerce shaped: inventory consistency under concurrent orders, cart and checkout idempotency, search and ranking, and handling a flash-sale traffic spike an order of magnitude above baseline.
90 min
Working, runnable code with clean class design. The heaviest round.
45 min
Defend and extend what you built. New requirement, how much do you rewrite?
45-60 min
Medium to hard, with complexity reasoning.
60 min
Inventory, checkout, search, or flash-sale traffic handling.
45 min
Ownership, scope and fit for the specific team.
Build the same handful of problems repeatedly until structure is automatic: parking lot, splitwise, elevator, inventory, ride matching. Spend the first ten minutes on entities and interfaces - working code with clean boundaries beats more features with tangled ones.
Yes. Runnable code with a main method or tests that demonstrate the flows is the baseline expectation. A design document with no working program scores poorly no matter how elegant the class diagram.
E-commerce problems from its own operations: inventory consistency under concurrent orders, cart and checkout idempotency, search and ranking, and surviving a flash sale that is ten times normal traffic.
Practise the rounds Flipkart actually runs, and get scored on where you stand.